Plastic Relaxation of Thermoelastic Stress in Aluminum/Ceramic Composites

C.T. KIM, J.K. LEE, and M.R. PLICHTA The dislocation generation due to a thermoelastic stress in 2024 Al/ceramic (SiC or TiC) composites was studied using transmission electron microscopy (TEM). Composites containing different ceramic particulates, ceramic volume fraction, and particle size were investigated. Dislocation density profiles were measured as a function of the distance from an A1/ceramic interface and compared with those calculated from an elastoplasticity model which accounts for the volume fraction of the ceramic particles. The intensity of dislocation generation showed a strong particle size dependence: as the ceramic particle size became of the order of a micron, the intensity of dislocation generation increased significantly. With an increase in the volume fraction of the ceramic particles, the dislocation density also increased, and the dislocation structure became a more tangled arrangement. If heat dissipation was taken into account as part of the plastic work, the predicted dislocation densities of the elastoplasticity model were found to be in reasonable agreement with the measured dislocation densities of 109 to 101~ c m -2. I. INTRODUCTION herent, it is expected that the dislocation generation due to a thermal expansion mismatch would depend strongly on the size of the ceramic particles in MMC's. Although the continuum mechanics and dislocation loop-punching models appear to yield fairly reasonable predictions for the experimentally observed dislocation generation, to date there have been no systematic experimental studies on the particle size-dependent dislocation generation. Additionally, the previous theoretical models were based on a single inclusion embedded in an infinite matrix, thus neglecting the effect of a multiparticle distribution. Therefore, a theoretical model for the plastic relaxation is desired to account for the volume fraction of ceramic particles in composites. The purpose of this work was to examine experimentally the dislocation generation due to the different thermal contractions in A1/ceramic (SiC or TiC) composites. Materials containing different ceramic particulates, volume fractions of the ceramic, and sizes of ceramic were investigated. An effort was also made to develop a theoretical model for the plastic relaxation around a spherical particle, which incorporates the volume fraction of ceramic particles. II. THEORY

P L A S T I C relaxation of the elastic stress and strain energy associated with both thermal misfitting inclusions and lattice mismatching precipitates has been long recognized. Examples of dislocation generation due to a thermal misfit in the vicinity of an inclusion include the observations by Vogelsang et al. t~l and Arsenault and Fished 2] on dislocation generation in the aluminum alloy matrix surrounding SiC whiskers or platelets and the finding by Chawla and Metzger ]3~ of higher dislocation densities at the C u / W composite interface. The mechanical properties of the composite, therefore, should be affected by the magnitude and extent of the dislocation generation that takes place in the soft matrix around hard particles as a result of the relaxation of thermal misfit stress in the interface region. The plastic deformation of a misfitting inclusion in an infinite matrix has been analyzed via a continuum mechanics model t4,5,61 as well as a dislocation looppunching model. 17,sl Lee et al. t41 were able to deduce a particle size-dependent yield stress from the AshbyJohnson model is] for the nucleation of a dislocation from a particle/matrix interface. They showed two effective yield stresses which depend upon the coherency of the particle/matrix interface. One is essentially independent of the particle size and equal to the theoretical yield strength, /z/2zr, where /x is the matrix shear modulus. The other effective yield stress applies to incoherent particles and is a strong function of particle size. When the particle is very small, say of the order of 10 to 20 nm, this stress approaches the theoretical yield strength. When the particle size becomes of the order of a micron, the effective yield stress approaches an average (lowtemperature) macroscopic yield stress. Since in metal/matrix ceramic composites (MMC's) the particle/matrix interface is considered to be incoC.T. KIM, Graduate Student, J.K. LEE, Professor, and M.R. PLICHTA, Associate Professor, are with the Department of Metallurgical Engineering, Michigan Technological University, Houghton, MI 49931. Manuscript submitted June 7, 1989. METALLURGICALTRANSACTIONSA

For mathematical simplicity, we assume that all of the ceramic particles are of spherical shape with radius a, and further, that a given particle is surrounded by uniformly distributed neighboring particles. This assumption allows us to use spherical coordinates whose origin is at the center of a ceramic particle with radial symmetry, as shown schematically in Figure 1. In the manner of Lee et al.,la] perfect plastic behavior is assumed for the matrix phase, and the crystallographic nature of plastic flow is neglected. However, the present model accounts for the volume fraction of ceramic particles. Thus, the matrix is considered to yield under the condition of a constant yield stress, and the yielding is considered independent of the stress axis. We shall first obtain the stress and strain associated with a misfitting spherical hard particle in the absence of plastic relaxation, i.e.,
VOLUME 21A, MARCH 1990--673

Substitution of the stress-strain Eq. [3] into Eq. [41 yields

1 d2

2 dr 2 (r2or) - or = 0
After integration, the radial stress is given by
B

[5]

or = A - --

F3

[6]

and from Eq. [1] the tangential stress is

B o0 = A + - -

2r 3

[71

Fig. 1 - - A schematic diagram for an idealized ceramic configuration and plastic zone.

The boundary conditions appropriate for small ceramic volume fractions are Or (r = a) = --Pl [8a] [8b] [8c]

under purely elastic conditions. Then, employing the foregoing assumptions and continuum plasticity theory, [9'101 the solution for the elastoplastic deformation is determined when plastic relaxation takes place in the matrix.

o'r (r = b) = - P 2 u (r = b) = a,,. A T . b

A. Pure Elastic State

Consider a thick hollow sphere whose internal surface of radius a is subjected to a pressure, P1, by the thermal misfitting ceramic particle and whose external surface of radius b is subjected to a pressure, P2, by the surrounding ceramic particles, as shown in Figure 1. Because of the radial symmetry approximation, the tangential displacements, as well as the shear stresses and shear strains, are all zero, and the radial displacement, u, is a function of the radial distance, r. Further, the equilibrium equations in the absence of body forces reduce to dot --+ 2(oro'0) -0 [11

Equation [8c] represents the displacement at the midpoint between two neighboring particles. Since the stress mode is critically influenced by the boundary conditions, their validity (especially Eqs. [8b] and [8c]) will be further discussed in Section V. Substitution of Eqs. [8a] and [8b] into Eqs. [6] and [7] leads to the stresses in terms of the contact pressures, P1 and P2, as

Pl a3 - P2 b3
O'r -b3 _ a3

a3b3(p1 - Pc)
(b 3 _ a3)r 3 [9a] [9b]

Px a3 - P2 b3
o0 = b3 _ a3

a3b3(p1 - P2) + 2(b3 _ a3)r 3

dr

Substituting Eqs. [9a] and [9b] into Eq. [3b] yields displacement as
u =

where or and o0 are radial and tangential stress components, respectively. The strains are related to the radial displacement, u, by

(1 + v m ) a 3 b 3 ( P 1 - P 2 ) 2Em(b 3 - a3)r 2

du e r - dr'

eo

u r

+ ("(1-2v~)(Pla3-P2b3) --~m(~ ~ a T

) + am'AT r
[10]

[21 Utilizing the boundary condition of Eq. [8c], P2 is given by 3a3(1 - vm)e 1
P2 = a3 + 2b 3 + V m ( a 3 _ 4b3)

For the third diagonal components, o 6 = o'0 and e 6 = e0 are implicitly assumed. H o o k e ' s law provides
1

[11]

er = E (Or
1

2v~176 + a" AT

[3a]

Further substitution of Eq. [11] into Eqs. [9a] and [9b] yields - P l a 3 ( 1 + V,n) Or = a3 + 2b 3 + um(a 3 _ 4b3) (1) - a3 + 2b 3 + lira(a3 _ 4b3) ~-~

eo = E [-V~ + (1 - v)O'o] + a" AT

[3b]

where ot is the thermal expansion coefficient and AT is the difference between the temperature of interest (room temperature in this study) and the solutionizing temperature for the composites. We note that with the above definition for the strain, the reference state is the one at the solutionizing temperature. The strain compatibility relationship is e~ = e0 + - 674--VOLUME 21A, MARCH 1990

203b3(1-- 2Vm)__P2

[12a]

-pla3(1 + vm)
O0 = a3 + 2 b 3 + ltm(a 3 - 4b 3)

r deo dr

a3b3(1-2v,,)P,
[4]
+ a3 ; ~ ; ;m~a3 ~4b3)

(1)
~

[12b]

METALLURGICALTRANSACTIONS A

Since both the constrained displacement and the traction must be continuous at the ceramic/matrix interc = eo m from face, r = a, we first obtain, by setting eo Eq. [3b], ar AT + - - [-vcor: + (1 - v~)or~]
l

in conjunction with the yield condition, or0 - orr = Ory, becomes dor~
dr

2ory
-= 0

[17]

e~

Integrating Eq. [17] and using the boundary condition or~= - e l a t r = a,

u~)or"~]

= a z " A T + - - [ - Uzor'r" + ( 1 -

Em

[13]

or~ = 2ory In ( r / a ) - Pl 0% = o'y[1 + 2 In ( r / a ) ] - e l ;

a <= r < rp

[18a] [18b]

C r m and by substituting 0-~ = or0 = -P~ = err into Eq. [13], we get p l [ b 3 - a 3 _ v , ( a 3 + 2b3)] atr= a

Yielding occurs when ore = ore at r = rp. Thus, replacing rp for a and Pp for P~ in Eq. [16], we obtain 3+ rp 2b 3 + Vm(r~ - 4b 3) = ory [191

or0 =

a3 + 2b 3 +

l~m(a3_

4 b 3)

Solving Eq. [13] for P~ yields

-(a., - ac)AT
P1
1

1-2 c
+ (1 - u,,) a3 + 2b 3 + Um(a 3 _ 4b3)_] + - - E c

114]

Substitution of Eq. [14] into Eqs. [12a] and [12b] provides the elastic stress components for the matrix phase. Note that if there is no difference between the two thermal expansion coefficients, am and ac, no elastic stress arises in the composite.

From Eq. [19],

ory[r3p + 2b ~ + Vm(r3p - 4b3)] pp =

3b3( 1 _ 2Urn)

[20]

By substituting Pp of Eq. [20] for P~ and rp for a in Eqs. [ 12a] and [ 12b], the stress components in the purely elastic region of the matrix are obtained as orr = -(1 3(1
--

B. Plastic Relaxation

When an effective elastic stress exceeds the yield stress of a metallic crystal, the crystal will undergo plastic deformation rather than remain in a purely elastic state. Since the stress state within the spherical ceramic is hydrostatic, the ceramic is considered to remain in a purely elastic state. Plastic deformation, therefore, can be considered to occur only in the matrix adjacent to the ceramic, as schematically shown in Figure 1, where rp is the radius of the plastic zone. For mathematical simplicity, two additional assumptions are made for the model on the plastic behavior of the matrix. First, we assume that the flow stress-strain behavior of the matrix phase is independent of the strain rate and also of the stress orientation. Second, we neglect strain hardening, so that the matrix phase is taken to be a perfectly plastic material whose flow stress-strain curve is horizontal at the yield stress, O-y. We adopt the von Mises yielding criterion, tlu namely, that yielding occurs when an equivalent stress, ore, exceeds the yield stress, o-y, where ore is given by
ore = ~
1

- ( 1 + Vm)O'y( r p / b ) 3 2ory 3(1 - 2Vm) -- ~ (rp/r) 3

[21a]
< b
r --

or0 -

2Vm)

1,'rn)Ory (rp/b) 3 +

OrY

(rp/r) 3 ; rp

<
=

[21bl

Since the radial stress must be continuous at the plastic front, the plastic zone radius, rp, is obtained by equating Eqs. [18a] and [21a] at r = rp.

2ory In ( r p / a ) - P, - - ( 1 +__12m)ory (rp/b) 3

3(1 2Vm)

2ory
3 [221

By applying the relationship between the displacement and the radial stress in the purely elastic state given in Eqs. [2] and [3b] to the radial stress given in Eq. [21a], the displacement outside the plastic zone is found to be

(1 + r'm)oryr3
3 E,~r 2

((1 + ~'m)ory(rp/b)30tm. AT) r

3 E,,

[(O'r -- 0"0) 2 31- (or0 -- 0"4') 2 "q- (orda -- orr)2] 1/2

[23] Substituting Eq. [23] into Eq. [2],

--(1 + Um)O'yr3(~ er = 3 E,,
-

[15]
Substitution of Eq. [12] into Eq. [15] yields
3a3b3(l ~ 2um)__P1 (1) or, = a3 + 2b 3 + Vm(a 3 _ 4b3) 7

~) + + am" A T

[24a1
rp<-r<-b

[16]

eo

3Era

+ a m ' A T ;

The equilibrium equation (Eq. [1]) for the plastic shell,

METALLURGICAL TRANSACTIONS A

[24b1
VOLUME 21A, MARCH 1990--675

Within the plastic zone, the strains are the sum of the plastic and elastic strains. Since the elastic strains are related to stresses by Hooke's law, we can write
du e~
. . . .

1
E m (0" r - - 2 t ' m t r O ) + a m " A T

dr

+ eer

[25]

In order to determine the equilibrium pressure, P1, upon plastic relaxation, use is made of the continuity condition for both the (constrained) displacement and the traction at the ceramic/matrix interface. From Eq. [3b], for the tangential elastic strain within the ceramic, we have
u 1

e0 . . . .

e0 . . . . [-Vmtr~ + (1 - Vm)tro] r Em
+ a m 9AT

Ec

[-Vctr~ + (1 - Vc)tr;] + a~. AT

-P1(1 - 2v~) [26]

-

+ ePo;

a <-- r < rp

E~

+ ac- AT

[31]

Here, e p and e p denote plastic strain components in the plastic zone. In order to obtain the displacement, u, in the plastic zone, we use the incompressibility condition for plastic strains, d + 2eP0 = 0. Multiplying Eq. [26] by 2, adding the result to Eq. [251, and using Eqs. [18a] and [ 18b] yields
- -

The displacement within the ceramic is then given by I - P , ( 1 - 2v~)

u = I E;

]
+ ac" AT r; 0 =< r =< a [32]

Finally, equating Eq. [29] and Eq. [32] at r = a provides a solution for P~:
(am - a ~ ) A T + (1 - Pm)try (rp/a) 3

du dr

2u r

2%(1 - 2Vm) [1 + 3 In ( r / a ) ] Em 3(1 - 2Vm)P~ + 3am'AT [271 Pl =

Em
1 -- 2Vm 1 -- 2vc

[33]

Em
for which the general solution is u= 2(1
-

Em

E~

2b'm)try

Em
Em

r In ( r / a )
.AT
a m

(1 - 2v~)Pi

r +
--"

< a --

<

rp

Substituting Eq. [33] into Eq. [22] for Pj gives an equation for rp which can be solved for a given set of material parameters, E, v, try, a, and AT. If the materials are elastically homogeneous, i.e., E m = E~ and Vm = V~, equating Eq. [29] to Eq. [32] with r = a yields directly
rp
=

r 2'

a[(ac - am) A T E m / ( 1

l.Pm)O'y] 1/3

[34]

[281 where C is an integration constant. Equating Eq. [23] to Eq. [28] with r = r p and rearranging the result with Eq. [22] furnishes an expression for u in Eq. [28]: u = 2(1
-

Further substitution of Eq. [34] into Eq. [22] yields P~ for an elastically homogeneous case.
C. Strain Energy

2Vm)Ory

Em

r In ( r / a )

The elastic strain energy inside the spherical ceramic, toc, per unit volume of the ceramic is given as
1 3P~(1 - 2vc)

toe = 2 [3trr(er - ac" AT)] -

2Ee

[351

(1 - ~m)tr,
+
Em

?,
r2'

a < r < rp
--

[29]

From Eqs. [25], [26], and [29], the strains in the plastic zone are given as
e r ~

where PI is given in Eq. [33]. The total work consumed in the plastic zone is the sum of the plastic strain energy and the elastic strain energy. The plastic work per unit volume of an element located at distance, r, is, by definition rl~

2(1 - 2Vm)try [In ( r / a ) + 1] + am AT

,

Em
(1 - 2vm)P~ 2(1 - Vm)% (rp/r) 3 [30a]
=

fo

trr dePr +

foeo

2tro deeo

[36]

Em
e0 =

Em
In ( r / a ) +
a m "

2(1 - 2Vm)O'y
Em

AT

(1 - 2vm)P~

(1 - v,.)try
+

Em
Em

em

(rp/r)

[30b]

where the usual tensor suffix notation is used in the first equation. From the incompressibility condition, e p + 2e g = 0 for the plastic strains, we have de~ = -2dePo. Substituting this relationship into Eq. [361 and noting that trr - tr0 = -try = a constant, we obtain
toP(r) =

e~ = -2eP0 = 2(1 - vm)try [1 - (rp/r)3];

a <- r < rp
--

f(
_

(Or -- 0"0) de?r

2(1 - 2v,,)try z [30c]

-OryerP

Em

[(rp/r)

3-

11

[371

676--VOLUME

21A, M A R C H 1990

METALLURGICAL TRANSACTIONS A

The elastic work per unit volume of an element located at distance r is

1

Table I. Some Values of Interparticle Spacing (Equation [40])

ek
+ 2o'0(e0 - e p - t~m" AT)] 0.001 0.002 0.005 0.010 0.020 0.050 0.100 0.200 0.500

b/a
4.497 3.591 2.686 2.183 1.802 1.450 1.275 1.160 1.073

we(r) = 2 [Orr(er -- ep -- O/m"AT)

6(1 - 2vm)tr2 [In (r/a)] 2 Em 2(1 - 2Urn) (2try -- 3P0try In (r/a)

Em
12(1 - 2Vm)O'yP~ The samples were solutionized at 490 ~ for 1 hour in a salt bath and subsequently water quenched. Dislocation structures were analyzed for both the composites and the control alloy using a PHILIPS* EM301 *PHILIPS is a trademark of Philips Instruments Corporation, Mahwah, NJ. transmission electron microscope operating at 100 kV. Transmission electron microscopy samples were chemically thinned with a 10 pct NaOH, 90 pct H20 solution to a thickness of less than 150/xm to prevent introduction of dislocations during a thinning process. Then, 3-mm-diameter disks were prepared and jet polished with a 30 pct nitric acid, 70 pct methanol solution using a potential of 35 V at a temperature of about - 3 0 ~ The dislocation density, p, was determined from a micrograph using the formula, p = 2N/Lt, where N is the number of dislocation intersections with a grid line of length, L, divided by the magnification, and t is the thickness of the sample, c~31 It should be noted that the dislocation density thus obtained can be lower than an actual value because of a low operating power of the EM301 microscope. The thin foil thickness was measured using a JEOL-100CX transmission electron microscope operating at 120 kV. Without liquid N2 cold stage, the foil surface was contaminated in 5 to 10 minutes by the focused electron beam. The thickness was measured from a micrograph showing the projected length between the contaminated spots of upper and lower foil surfaces using t = d/sin 0, where d is the projected length and 0 is the primary tilting angle, equal to ~ 4 0 deg in this study. The thin foil thickness was measured as a function of distance from the A1/ceramic interface. Each thickness value at each point was incorporated for measurements. A circular test line was used to eliminate the effect of any preferred orientation of a dislocation structure, and the length of a grid line was 76 cm.

- (1 - v,.)O'y 2 - ~ (1 - 2v..)P~

] [381

Similarly, the elastic strain energy per unit volume stored outside the plastic zone is toOUt(r) -(1 + Yr.) Ory 6(1 - 2vm)E~

(rp/b) 6 +

(1 +

(rp/r) 6
[39]

3E,,

D. Relationship between 2b and Volume Fraction, 4)

It is now necessary to find a relationship between the ceramic volume fraction, ~b, and interparticle distance, 2b. For a polydispersed system of particles, an accurate relationship is difficult to obtain. However, when all of the particles are assumed to have the same radius, a, it can be shown f121 that for randomly distributed particles, the average half distance, b, between two neighboring particles is b = a + a" exp (8~b) _I X-2/3e-Xdx 6q~l/3 ..~,

[40]

Equation [40] is known to be accurate when the volume fraction, ~b, approaches zero. Thus, the maximum ~b value in this study is limited to 0.1. Table I lists'some b/a ratios obtained with Eq. [40].

III.

EXPERIMENTAL PROCEDURES

Composite materials of the prealloyed 2024 Al powders and ceramic particles were fabricated via hot pressing and hot rolling. The average sizes of the ceramic particles are listed in Table II. The ceramic materials were blended with prealloyed 2024 A1 powders, vacuum hot pressed at 495 ~ for 2 hours as 50-mm-diameter, 13-mm-long disks under a pressure of 3000 psi. The hot-pressed samples were then wrapped with thin A1 foils in order to prevent any reaction between the composites and the copper tube during heating and hot rolling, encapsulated with copper tube under vacuum, heat-treated to 515 ~ (above the solidus temperature), and hot rolled to 2-mm-thick sheets in five passes. Three different volume percents of ceramic particles were considered: 0, 2, and 10 pct. The 0 vol pct sample served as a control. METALLURGICALTRANSACTIONSA

Table II. Average Size and Volume Fraction of Materials Matrix 2024 2024 2024 2024 AI AI Al AI Ceramic SiC SiC SiC TiC (22/.tm) (14/zm) (5/tm) (30/zm) Vol Pct 0, 0, 0, 0, 2, 2, 2, 2, 10 10 10 10

VOLUME21A, MARCH1990--677

IV.

RESULTS
A,'2024 -- SiC

A. Stress and Strain Energy Distribution in Composites

Material parameters used for the calculation are given in Table III. Figure 2 presents the normalized stress with respect to the yield stress of the matrix as a function of the normalized distance from the ceramic center in the AI/SiC system. The value of unity for r / a corresponds to the matrix/ceramic interface. The stress distributions in the purely elastic state and in the elastoplastic state are displayed for ~b = 0 (a single particle) and ~b = 0.003, respectively. Matrix stress components, o-~ and tr0, ate relatively short-ranged, decaying effectively to zero at approximately r -- 3a. It is noted that the difference in the stress state between the two cases is significant within the ceramic particle as well as in the plastic zone but not in the elastic region of the matrix. Inside the plastic zone, the tangential stress is quite different, with the sign being reversed from its counterpart of the pure elastic case. The plastic zone size is seen to increase with increasing ceramic volume fraction. In Figure 3, the normalized strain energy density is plotted as a function of the normalized radial distance for the three different volume fractions. The strain energies were normalized with respect to the plastic strain energy per unit volume for ~b = 0 at r = a. When the plastic zones overlapped each other, the plastic strain energy densities were calculated through superposition. Note that the plastic strain energy density decreases inversely with r 3 (Eq. [36]) and is seen to increase with volume fraction of ceramic. In the case of ~b = 0, the theoretical plastic zone size, rp, is found to be 2.41a for A1/SiC composites and 2.27a for A1/TiC composites. The theoretical plastic zone size, rp, is found not to increase rapidly with an increase in ~b, but plastic zones start to overlap each other at ~b = 0.007 for A1/SiC composites as rp ~- b at this volume fraction.
B. Dislocation Generation

.++
9

-1C

/ // 9 // / / // //

// /;,~

- +-+:0003):0 - -

pure

elastic

9 9

0 ) elasto- plastic 0.003

//

-1,'

+
r/a

Fig. 2 - - T h e stress distribution in the purely elastic state and in the elastoplastic state in A1/SiC composites. The matrix/ceramic interface locates at r/a = 1. The volume fraction effects are displayed for ~b = 0.003. The stresses are normalized with respect to try.

Similar to the findings of Vogelsang et al.,m the intensity of dislocation generation at the A1/ceramic interface is related to the size and shape of SiC particles. Figure 6 reveals a low intensity for dislocation generation in the vicinity of a small SiC particle with diameter

A
L.

A12024/SIC ....

ttt~

- -

elastic plastic

2[ 1 ..... 3 ......

o
1:0.0

Dislocation density gradient showing a higher density near the A1/SiC interfaces is observed in the as-quenched ~b = 0.02 and 0.1 A1/SiC composites (Figures 4 and 5). Line, helical, and loop-type dislocations can be seen in the matrix. For A1/SiC composites with ~b = 0.02, the dislocations near the A1/SiC interface are generally in tangled arrangements, whereas dislocations away from the SiC particle are not, as shown in Figure 4. As the SiC volume fraction increased, dislocation density seemed to increase, and the dislocations also occupied a wider area (Figure 5). As expected, more dislocations were generated at a comer of a SiC particle, where a higher stress was concentrated during cooling (arrow marked).

2:0.02
3:0.1

L2
Poisson's Ratio 1/3 0.19 0.19

2
r/a

r?

"

Fig. 3 - - T h e strain energy distribution in the elastoplastic state in A I / SiC composites. The strain energies are normalized with respect to the plastic strain energy per unit volume at r = a for the case of ~b = 0. The volume fraction effects are displayed for q~ = 0.02 and 0.1.

Table III.

Material Parameters

Material 2024 A1 SiC TiC

Yield Strength (MPa) 76 ---

Elastic Modulus (GPa) 73 510 310

Thermal Expansion Coefficient (K -a) 24 X 1 0 - 6 4.4 X 1 0 - 6 7.4 X 10 -6

Thermal Misfit Strain

--

0.0091 0.0077

678--VOLUME 21A, MARCH 1990

METALLURGICAL TRANSACTIONS A

Fig. 4 - - D i s l o c a t i o n generation near an A1/SiC interface for to = 0.02. The ceramic particle diameter is 7.4 ~ m . Note the small intermetallic compounds associated with a 2024 A1 alloy.

Fig. 6 - - D i s l o c a t i o n generation in the vicinity of a small SiC particle for to = 0.02. The ceramic diameter is 0 . 8 / z m .

0.8/xm. No significant dislocation density gradient is observable in this case. As the particle diameter increases (to - 7 / x m in Figure 4 and --17/xm in Figure 7), the intensity of dislocation generation increases significantly. In comparing the present result with those of Vogelsang et al. tl] and Arsenault and Fisher, t2] an interesting observation is that much more helical dislocations seemed to be generated in the 2024 A1 matrix than in the 1100 AI or 6061 A1 matrices. No detailed studies were performed to find the cause of this morphological difference, but one possible reason is that for the present 2024 A1 alloy, the water quenching right after a solutionizing treatment should have produced quenched-in vacancies or vacancy loops, which might have helped to nucleate helical dislocations. The 1100 and 6061 A1 alloy samples, however, were furnace- or air-cooled after annealing. [1,2] Figure 8 shows dislocation generation in a ~b = 0.02 A1/ TiC composite, in which the particle size is 16.2/xm. As compared with an AI/SiC composite in Figures 4 or

7, the dislocation density is found to be significantly less. This decrease can be attributed to smaller differences in both the thermal expansion coefficient and elastic modulus between A1 and TiC (Table III). Dislocation density is plotted as a function of a normalized distance from the A1/ceramic interface in Figure 9. Also included are theoretical density curves obtained with the plastic strain energy term of Eq. [37]. Theoretical dislocation density is approximated as Pt -~ Po + f a r ~ where/90 is the dislocation density in the control alloy ( - 5 x 108 cm-2), /z is the matrix shear modulus, b is the Burgers vector, and f is a partitioning parameter which reflects the fractional amount of the plastic strain energy stored as dislocations. In plastic deformation, part of the plastic work is dissipated, and defects other than dislocations are produced. However, in the absence of our detailed knowledge on the partitioning of a plastic work, the parameter, f, was arbitrarily taken as 0.5 in Figure 9. Each set of the experimental data in Figure 9 represents density measurements from a TEM micrograph. V. DISCUSSION

Residual stress is an inherent characteristic for composite materials. When composites are cooled to the room

Fig. 5--Dislocation generation near an M / S i C interface for tO = 0.1. The ceramic particle diameter is 7.4 ~ m . METALLURGICAL TRANSACTIONS A

Fig. 7 - - D i s l o c a t i o n generation in the vicinity of a large SiC particle for to = 0.02. The ceramic diameter is 17.4/zm. VOLUME 21A, MARCH 1990--679

Fig. 8 - - D i s l o c a t i o n generation in an A1/TiC composite with ~b = 0.02. The particle size is 1 6 . 2 / ~ m .

temperature from the fabrication or solutionizing temperature, residual stresses are introduced into the composites due to a mismatch in thermal expansion coefficient between the matrix and ceramic phases. Ledbetter and Austin I~5~ measured the average residual stress for 6061 AI composites containing 30 vol pet SiC particles via X-ray diffraction analysis. A hydrostatic tensile stress was found in the aluminum matrix, while a hydrostatic compressive stress was found in the SiC particles. Arsenault and Taya tt61 observed that the tensile yield stress was greater than the compressive yield stress by - 1 3 MPa for 6061 A1 composites containing 20 vol pet spherical SiC particles, suggesting a possibility that a

20

sic

Vol.%
p

a(prn)

E u ,=
0

15
TiC

9 9 9 z] o +

1~ ) 10 10 2 2 2 2

theoretical
11.6 3.7 8.7 3.7 0.4 8.1

0
" 5

01

1.0

Control Alloy J I-"

"

+ a

to"

13 a

1.5
r/a

2.0

Fig. 9 - - D i s l o c a t i o n density profile as a function o f a n o r m a l i z e d distance f r o m an A l / c e r a m i c interface. The theoretically c a l c u l a t e d densities are d i s p l a y e d for ~ = 0 . 0 2 and 0.1 A I / S i C c o m p o s i t e s .

compressive residual stress existed in the matrix. Though at first glance the two measurements appear to contradict each other, we should note that residual stresses depend on several factors, such as the volume fraction and distribution of ceramic particles f~Tjand also possibly on the details of heat transfer during cooling. When the ceramic volume fraction is small, we may assume that each ceramic particle is surrounded by the matrix phase, and the average interparticle spacing is not influenced by the presence of a ceramic phase with a lower thermal expansion coefficient. In this situation, spherical ceramic particles are in a hydrostatic state and their surrounding matrix phase suffers a compressive radial stress state. Thus, on the average, the composites should produce a compressive residual stress. On the other hand, when the volume fraction is so high that the ceramic particles touch each other, the ductile matrix phase should be considered enclosed by the ceramic phase. In this case, the ceramic particles are still in a compressive stress state, as is the matrix in the proximity of the ceramic particles. However, the majority of the matrix phase with a higher thermal expansion coefficient should undergo shrinking on cooling, resulting in a tensile residual stress. The actual conditions of composites may fall between the two extreme cases. In light of this discussion, we note that the boundary conditions (Eqs. [8a] through [8c]) employed in the present model are based on the approximation of a small volume fraction, ~b, and thus the model's prediction should yield an overestimation for the plastic strain energy density (and the dislocation density) at a high q~. In this regard, other boundary conditions reflecting the volume fraction of ceramic particles are certainly desirable as a future study. According to the early studies on the inclusion sizedependent yield stress, I4'7'8! the effective yield stress for dislocation generation should approach the macroscopic yield stress of the matrix phase if the ceramic particle size becomes of the order of a few microns and the ceramic-matrix interface is incoherent. Therefore, the dislocation density gradient showing a higher density near the A1/ceramic interface should be observed when the ceramic particle size is over a few microns (Figures 4, 5, and 7). No significant dislocation density gradient, however, is observed when the particle diameter is less than 1 /~m (Figure 6). Several reasons can be offered for this case. The stress due to the thermal misfit could be less than the effective yield stress for dislocation generation, t4,8] Alternatively, in case the plastic zone size is smaller than the characteristic dislocation punching distance, the dislocation generated fn'st would possibly reach the plastic zone front, thus limiting the further dislocation generation through the influence of the back stressfl 1 It should be also noted that a stable, polygonized dislocation structure can be developed around a submicron particle, as nicely demonstrated in the high-voltage electron microscopy work of Vogelsang et al. t~] Although Figure 9 shows a fair agreement, at least qualitatively, between the measured and calculated dislocation density, we note that the calculated dislocation density depends on the partitioning parameter, f, that appeared in the formula p, ~ Po + ftoP(r)/(O.5txb2). As mentioned before, some of a plastic work is dissipated

6 8 0 - - VOLUME 21A, MARCH 1990

METALLURGICAL TRANSACTIONS A

as heat and some is spent for producing various defects. One can estimate the value o f f from a macroscopic tensile or compression test. However, it should be noted that the present microscopic situation is quite different from a macroscopic case; thus, the use of such an experimental value for f is not necessarily valid. In view of this uncertainty o f f , further studies are necessary for a useful application of the present micromechanics analysis. As an alternative, an effective plastic strain can be used for estimating dislocation densities, but it requires a mean free path for dislocations, which is again a quantity as ambiguous as heat dissipation. ~ When two ceramic particles with a thermal misfit strain approach each other, there arises an additional stress field due to the difference in elastic constants between the matrix and ceramic phase. 04,~81 However, the strength of this inhomogeneity-induced stress is proportional to r -6 (as compared with the r -s terms of Eqs. [12a] and [12b]) and was neglected in the present analysis. No in situ TEM work was performed for the present 2024 AI composites, but the work of Vogelsang et al. t~ on 6061 A1/SiC composites found that most of the dislocations were annealed out at 507 ~ Thus, it appears quite reasonable to use the solutionizing temperature of 490 ~ as the reference temperature at which all of the displacements were assumed to disappear. The plastic relaxation of the matrix surrounding a thermal misfitting ceramic particle can lead to substantial changes in the associated stress field and strain energy. This results in elastic and plastic strain energy gradients around the ceramic particles. Aluminum alloy matrix/ceramic composites are somewhat unique since the properties of the aluminum matrix itself can be modified and varied significantly through the precipitation process. The drastic change in the stress field surrounding the ceramic particles, therefore, can affect diffusion, kinetics of precipitate nucleation and growth, coherency state, etc. It would be of great value to understand the effects of the plastic relaxation on the precipitation kinetics.

NOMENCLATURE
a

b (rr (re (re (ry

er eo

e~
ePo E
v ol

T AT
u

rp P1 P2

4,

spherical ceramic radius half of interparticle spacing radial stress component tangential stress component equivalent stress yield stress of the matrix phase radial strain component tangential strain component radial plastic strain component tangential plastic strain component elastic modulus shear modulus Poisson' s ratio coefficient of thermal expansion temperature temperature of measurement - solutionizing temperature radial displacement plastic zone radius internal pressure external pressure volume fraction of the ceramic

Subscripts m c matrix ceramic

ACKNOWLEDGMENTS The authors are grateful to Professor R.J. Arsenault for his critical reading of the manuscript and many constluctive suggestions. They also wish to acknowledge the financial support provided by ALCOA (contributions of CTK and MRP) and by the Division of Materials Research of the National Science Foundation under Grant No. DMR-8508720 (contributions of JKL and MRP).

REFERENCES
VI. CONCLUSIONS
1. M. Vogelsang, R.J. Arsenault, and R.M. Fisher:. Metall. Trans. A, 1986, vol. 17A, pp. 379-89. 2. R.J. Arsenault and R.M. Fisher: Scripta Metall., 1983, vol. 17, pp. 67-71. 3. K.K. Chawla and M. Metzger: J. Mater. Sci., 1972, rot. 7, pp. 34-39. 4. J.K. Lee, Y.Y. Earmme, H.I. Aaronson, and K.C. Russell: Metall. Trans. A, 1980, vol. l l A , pp. 1837-47. 5. C.A. Hoffman: J. Eng. Mater. Technol., 1973, vol. 95, pp. 55-62. 6. Y.Y. Earmme, W.C. Johnson, and J.K. Lee: Metall. Trans. A, 1981, vol. 12A, pp. 1521-30. 7. W.C. Johnson and J.K. Lee: Acta Metall., 1983, vol. 31, pp. 1033-45. 8. M.F. Ashby and L. Johnson: Phil. Mag., 1969, vol. 20, pp. 1009-22. 9. R. Hill: The Mathematical Theory of Plasticity, Oxford University Press, London, 1950, p. 97. 10. A. Mendelson: Plastici~-- Theory and Applications, MacMillan Publishing Company, New York, NY, 1968, p. 135. 11. G.E. Dieter: Mechanical Metallurgy, McGraw-Hill, New York, NY, 1961, p. 58. 12. A.J. Ardell: Acta Metall., 1972, vol. 20, pp. 61-71.

From the experimental data and the theoretical model, the following conclusions were made: 1. A theoretical model for the plastic relaxation of thermal misfit stress in composites was developed such that the model incorporates the volume fraction of spherical ceramic particles. Its prediction of the plastic strain energy gradient around the ceramic particles was found to be in a fair agreement with the observed dislocation density gradient. 2. As the volume fraction of ceramic particles increased, the dislocation density increased, and the dislocation structure became more tangled due to the interaction between thermal misfit stress fields. 3. The intensity of dislocation generation due to the different thermal contractions seemed to be strongly influenced by the ceramic particle size and also by the difference in elastic modulus between the ceramic and matrix.

METALLURGICAL TRANSACTIONS A

VOLUME 21A, MARCH 1990--681

13. P.B. Hirsch, A. Howie, R.B. Nicholson, D.W. Whelan, and D.W. Pashley: Electron Microscopy of Thin Crystals, 2nd ed., Krieger, New York, NY, 1977, p. 422. 14. J.D. Eshelby: Proc. R. Soc., 1957, vol. 241A, pp. 376-96. 15. H.M. Ledbetter and M.W. Austin: Mater. Sci. Eng., 1987, vol. 89, pp. 53-61.

16. R.J. Arsenault and M. Taya: Acta Metall., 1987, vol. 35, pp. 651-59. 17. N.K. Asamoach and W.G. Wood: J. Strain Anal., 1970, vol. 5, pp. 88-97. 18. W.C. Johnson and J.K. Lee: Metall. Trans. A, 1979, vol. IOA, pp. 1141-49.

682--VOLUME 21A, MARCH 1990

METALLURGICAL TRANSACTIONS A